Slashdot Mirror

← Back to Stories (view on slashdot.org)

Fighting Android Sparring Partner

Posted by Zonk on from the so-many-ways-this-could-go-badly dept.

GeneOff writes "Just in time for Christmas, comes FA1, the fighting android from SDT (Self Defence Technologys.) I loved my Rock-Em, Sock-Em robots I got one year from Santa. But it was tough getting opponents. Well, no more. Here is a hackable real robot that won't whimp out on you." From the article: "... a robot that can jab, hook, and cross, but still keep the violence to a minimum with adjustable difficulty levels. The FA1 can also dodge your punches with "human-like" movements and he should be hanging out on the show floor at CES -- so we'll be sure to challenge him to a round of verbal barbs from a distance."

3 of 111 comments (clear)

  1. Doubtful by kai.chan · · Score: 4, Insightful

    There is a few reasons why I doubt the usefulness of this robot.

    1) To develop an "android" with "human-like" combative movements is unlikely at this point in time. The Actroid shown in the videos here is quite advanced in mimicing human motions, but the movements are very jerky and the motor control is very obvious.

    2) The previous point leads to another problem. Training with a mechnical dummy is detrimental to gaining actual combat experience. The "android" will , no doubt, move in a series of jerky movements predefined by state machines. A human fighter flows from one move to another. The android can't help replicate real combat unless it has physics and advanced algorithms programmed into it. Also, the android shown in the picture seems to be stationary. Boxing and martial arts is very dynamic, so the android serves more like a game than a serious combat trainer.

    3) The article claims that the android can dodge punches. I say that it can't. It takes a fist less than a quarter of a second to travel from the ready position to the opponent's face. The microprocessor needs to do image processing to realize that the fist is moving towards a specific area, which will require many image frames before it can actually determine the fist movements. And even if the microprocessor was able to act fast enough, I haven't seen any motors that can rotate fast enough.

    4) Cost. A stepping side-kick has tremedous amount of force. If a kick can knock a human 10 feet back, I doubt that the android can sustain its working condition after numerous attacks. When it breaks down, how much would it cost to get it serviced?

    1. Re:Doubtful by gordo3000 · · Score: 2, Insightful

      While I agree with your first 3 points, I don't think the fourth is a valid problem.

      The danger of the side kick is not that it can't be dodged easily, but that when it hits, it can do a lot of damage to a person. But that is the key, it is to a person. This is a man made machine. It can easily be built to withstand even the strongest kicks. People aren't all that strong in the grand scheme of things, it is only when you compare people to other people that the strength seems incredible.

      of course, I have no idea what they mean by dodging. The processor speed and image decoding are just computational problems and easily overcome with very expensive hardware. I find the real problem in having a machine beable to move in teh number of directions necesarry. Anyone who has done any type of boxing/martial arts nkows that if you always respond the same way (or some small set of ways) to one attack, it is trivial to lure the person into a trap. The lack of dynamic and varied decision making would be the biggest downfall.

      either way,it would be a cool toy. and for a lot of people, he will probably be challenging.

    2. Re:Doubtful by tamnir · · Score: 2, Insightful
      3) The article claims that the android can dodge punches. I say that it can't. It takes a fist less than a quarter of a second to travel from the ready position to the opponent's face. The microprocessor needs to do image processing to realize that the fist is moving towards a specific area, which will require many image frames before it can actually determine the fist movements. And even if the microprocessor was able to act fast enough, I haven't seen any motors that can rotate fast enough.

      It looks like you missed that story from June, about robotic hands catching fast flying stuff. The required image processing power and fast motors are indeed available.

      I do agree with you though, and also call BS, but just had to point out that your 3rd point is flawed.
      --
      I code, therefore I am.